Plan/Act model usage patterns in Cline

We analyzed usage patterns across Plan and Act modes in Cline over the last 7 days. The data reveals clear preferences in model selection and interesting patterns in how developers combine different models for planning versus execution.

Plan mode usage distribution

Claude Sonnet 4 dominates planning with 42.6% of all Plan mode usage, nearly triple the share of second-place Gemini 2.5 Pro at 15.3%.

Act mode usage distribution

Execution patterns show even stronger concentration, with Claude Sonnet 4 handling 46.6% of all Act mode tasks. We expect Sonnet 4.5 to replace Sonnet 4's usage over the coming weeks for Acting as well as Planning.

Cross-mode pairings

When developers choose different models for Plan vs Act, specific patterns emerge. The Claude Opus 4.1 → Claude Sonnet 4 combination leads at 25.3% of all cross-mode usage.

Model pricing context

Understanding these patterns requires considering the economics. Here's the current OpenRouter pricing (per million tokens):

Claude models:

• Claude Sonnet 4: $3 input / $15 output (1M context)
• Claude Sonnet 4.5: $3 input / $15 output (1M context)
• Claude Opus 4.1: $15 input / $75 output (200K context)

Gemini models:

• Gemini 2.5 Pro: $1.25 input / $10 output (1M context)
• Gemini 2.5 Flash: $0.30 input / $2.50 output (1M context)

OpenAI models:

• GPT-5: $1.25 input / $10 output (400K context)
• GPT-5 Mini: $0.25 input / $2 output (400K context)

Other models:

• Grok Code Fast 1: $0.00 input / $0.00 output (256K context)
• DeepSeek Chat v3: $0.24 input / $0.84 output (164K context)
• DeepSeek R1: $0.40 input / $1.75 output (164K context)

Data notes

• Percentages based on usage patterns from the last 7 days
• Pricing data as of October 2025

For detailed model comparisons and configuration options, see the Model Selection Guide. Share your usage patterns on Reddit or Discord.